ULTRASONIC INSPECTION OF PIPELINE SPLIT-TEES

F.H. Dijkstra1, N. Prtzgen1, and H.A.M. van Merrinboer2

1
Rntgen Technische Dienst bv, Rotterdam, The Netherlands,
2
N.V. Nederlandse Gasunie, Deventer, The Netherlands

Abstract: For repair and hot-tapping of pipelines in the field, often so-called repair sleeves or split-tees are used.
These consist of two semi-circular pipe segments, joined by two longitudinal welds and attached to the pipe by two
circumferential fillet welds.
Whereas the Time of Flight Diffraction (ToFD) method is successfully applied for inspection of the longitudinal
welds, NDE possibilities for the circumferential fillet welds are limited. Radiography is not an option, a.o. because
of the thickness variations. The welds complex, varying geometry prevents a reliable volumetric ultrasonic
inspection.
N.V. Nederlandse Gasunie and Rntgen Technische Dienst bv in The Netherlands are conducting a modeling study
in order to optimize parameters for inspection by ultrasonic phased arrays and ToFD. The ultimate goal is to
achieve a level of NDE reliability for these welds that is comparable to that of circumferential butt welds. The
presentation will highlight some results and provide an outlook on the implications of split-tee fillet welding and
NDE.
Introduction: In The Netherlands, manual ultrasonic examination of circumferential fillet welds of assemblies
used for repairs and hot-taps on gas transport pipelines became mandatory early 2003. This decision was related to
failure of a split-tee fillet weld in November 2002. Before 2003, the examination requirement of these fillet welds
had been Magnetic Testing only.
Pipeline repairs and hot-taps are usually performed on gas transport pipelines during operation, meaning that the
pipeline is under full (gas) flow and pressure. The presence of the gas flow results in extremely high cooling rates
during welding. Special precautions are taken to ascertain that possible weld defects will not result in cracks that
propagate into the original pipe wall. These measures include optimized welding procedures and consumables (if
anything fails it will be the weld rather than the pipes parent metal). In addition, welders are especially trained for
this activity. Their skills are annually tested within a recertification scheme.
After manual UT became mandatory early 2003, an NDE procedure was developed covering the perpendicular
fusion line (between sleeve and weld) only, using standard pulse-echo UT probes (45 and 60) and an arbitrary
sensitivity setting using the DGS approach using a 2 mm dia flat bottomed hole as a reference reflector.
Performance level of this approach was evaluated by means of a Round Robin test covering four fillet welds and
five UT operators. In this exercise, not only the intended artificial defects were detected but also many typical fillet
weld-related small non-relevant imperfections were found. In practice, this result could lead to unnecessary repairs,
which could represent a safety risk by itself.
In order to improve the quality level of UT on split-tee fillet welds, a computer modeling study was initiated as a
joint effort of N.V. Nederlandse Gasunie and RTD bv. This study is performed in two stages:
modeling performance of UT based on conventional probes as a function of geometry, defect type/size and
probe angle; this part of the study can directly be compared to practical experiments performed on test welds,
thus validating the model;
using the model to optimize NDT parameters, in which exercise also phased arrays and Time of Flight
Diffraction techniques are included. This aims for a full volumetric examination.
This paper summarizes results of the modeling study, experiments and preliminary conclusions for future split-tee
welding and NDT. The typical geometry of a split-tee fillet weld is shown in fig. 1. Fig. 2 shows a split-tee during
welding in the field.

Fig. 1: Typical Split-tee in lab. Photograph: Gasunie.

Fig. 2: Split-tee during welding. Photograph: Gasunie.

In-service welding of pipelines: There can be two reasons for welding on pipelines under operating conditions.
Firstly, the need to make a branch connection by a split-tee or so called weldolet (hot tapping). Secondly to
repair a pipe that has been locally corroded or mechanically damaged by using a repair sleeve. What these two
constructions have in common is the presence of two circumferential fillet welds to connect the split tee or repair
sleeve to the pipeline.
Metallurgically, these welds are of main concern. The gas flow results in extremely high cooling rates, with hard
and brittle heat affected zones in the parent pipe metal as a consequence. In addition, the fillet weld configuration
itself is more crack sensitive compared to the relatively small pipe wall thicknesses, because of its typical shape
and constrained condition caused by the relatively heavy wall of the split tee or repair sleeve. Although pre- and
post weld heat treatment could be carried out to improve the weld's quality, these measures are impractical and
costly. Also, avoiding burn through, high heat inputs softening the pipes heat affected zone are not possible.
Despite all these challenges, Gasunie has successfully welded hundreds of hot tap and repair sleeves during the last
few decades, on pipeline material qualities varying between the old high carbon API X 52 (carbon equivalents up
to IIW 0,51) and the modern thermo-mechanically treated high-strength X70 materials. For instance, heat affected
zone hardening in old pipe material could excess almost 500 HV under rapid cooling and low heat input conditions,
nevertheless still weldable without any (cold) cracks! Successful welding under these extreme conditions was only
possible by especially trained welders, well-balanced weld procedures, sound weld bevel design and proper quality
control of all relevant parameters.
The success of the weld process itself is mainly featured by its especially developed weld consumable. Extreme
low hydrogen (< 2,5 ml/100 gr), low yield (< 425 N/mm2), low carbon (< 0,01 %) and high toughness are the most
important properties of the weld metal deposit, avoiding (delayed) cold cracking. In addition, temper bead welding
softens the pipe's heat affected zone to a more acceptable hardnesses in the range of 350-400 HV.
Despite these preventive measures, failure rate of hot tap and split-tee repair welding is in the range of 10-3 - 10-4.
This performance level may be sufficient in many cases, but increasing awareness of (societal) risks demands better
performances, equivalent to the acceptable failure rate of pipeline construction girth welding of about 10-5 - 10-6, at
least. Therefore, Gasunie aims to improve the integrity level of hot tap girth fillet welding with a factor of 100.
Complete volumetric inspection of the girth fillet weld, including heat affected zones at the weld toe, underbead
parent pipe metal zone and vertical fusion line should contribute to this enhanced performance level. In practice,
this means that the Probability Of Detection of a new volumetric inspection technique should achieve more than
90%, preferably higher. This challenge is the main drive to support the development of advanced NDT techniques
such as ToFD and Phased Array, to inspect these critical fillet welds.
The typical geometry of a split-tee or repair sleeve fillet weld is shown in fig. 3.

Fig. 3: Typical split-tee geometry

Description of the model: In the frame of another project "Phased Array Imaging", carried out by RTD bv in
cooperation with Delft University of Technology, a model has been developed that will be used to develop and
optimize imaging of weld defects using Phased Array technology [1]. This model is based on a finite difference
scheme, and was developed in a MathLab environment. Unlike other (acoustic) models, this model is an elastic one
and takes into account all possible reflected and mode converted signals that can occur during travel of an
ultrasonic wave front through solid material.
Of course, the fact that the model is two-dimensional prevents that defect length is taken into account. The model
assumes defects to be longer than the beam width in lateral direction.
The model enables implementation of any geometry in a two-dimensional plane. The object's boundaries can be
defined as "hard" (reflective) or "soft" (absorbing) for ultrasonic waves. Reflectors of any shape, dimension,
orientation and nature can be built in. Parameters such as ultrasonic frequency, resolution, element size, range and
defect characteristics can be varied. The model provides visualization of the moving wave front as well as energy
distribution in the beam and in the component. In addition, the A-scan is calculated, providing information on
signal amplitude, transit distance and signal shape. Measurements on given reflectors can be scaled to a reference
reflector, the same way a calibration prior to a real ultrasonic inspection is performed. Amplitude scaling can also
be performed off-line.
Once all parameters have been set, each simulation typically takes two hours on a 3 GHz Intel Pentium 4 PC with
1024 MB of memory. Series of simulations, for instance where a full scan over a defect is required between half
and full skip, the same way an ultrasonic operator would scan, take a few days to calculate. The position increment
in this case is typically 1 mm.
Simulation parameters used: In phase 1, where the primary objective was the evaluation of conventional UT
performance, the following parameters were varied:
Two typical geometries, within the boundaries of visually acceptance were selected: ideal and irregular;
Two typical sleeve thicknesses were chosen, 10 and 18 mm;
UT probe parameters were derived from standard angle transducers of 45, 60 and 70, 4 MHz;
Lack of fusion type defects in the fusion line between repair sleeve and weld were used, under 90 and 45, with
varying height, as well as underbead defects at the toe of the weld and slag-shaped defects.
In fig. 4, the situation and the primary parameter settings (starting conditions) for the simulations on the thinner
sleeve (10 mm) on the ideal geometry are shown. In fig. 5, the geometric parameters and defect types (boundary
conditions) are shown. For this figure, a macro photograph of a split-tee fillet weld was used. Varying all
combinations of these parameters, the total number of simulations amounted to approx. 200.

probe position geometry

underbead defect
probe angle

slag
fusion line defect

Fig. 4: Ultrasonic and material parameters

Fig. 5: Geometric parameters and defect types

Round-Robin tests: The results of phase 1 modeling were compared to the outcome of a series of inspections
(round robin test) on four representative repair sleeve fillet welds with sleeve thicknesses of 10 and 18 mm. These
welds contained multiple defects. The round robin test involved five UT level II ultrasonic operators, who
inspected the welds according to the same written procedure. Sensitivity was based on 2 mm diameter flat bottom
hole (DGS method). All welds were also examined with an optimized RT technique for reference purposes, using
single wall single image exposure (which is, in practice, not possible for obvious reasons). Destructive examination
of the welds is planned at a later stage, reason why the present results can only be preliminary.
Results of phase 1 modeling: From the modeling results, the following could be learned:
Indications of fusion defects at the vertical fusion line (between sleeve and weld) appear at a range where no
geometrical echoes are detected. This leads to a clear detection of these defects, not disturbed by geometry.
Because in a paper like this no moving images can be included, fig. 6 shows two captured images of the
simulated moving wave front. In fig. 6A, the wave front generated by the crystal is just about to hit the fusion
defect. Part of the ultrasonic energy has already reflected against the gap between sleeve and pipe wall. Blue is
compression waves in this figure, red is shear waves. In fig. 6B, the signal has just hit the crystal (blue wave
front just behind the crystal), reception has just taken place, and the A-scan has been detected with a good
signal to noise ratio.

Fig. 6A: Wave front is about to hit defect

Fig. 6B: Reflected wave front just hit crystal

On vertical fusion defects in the vertical fusion line, there is a good relationship between defect height and
amplitude for defects up to approx. 4 mm. Above that height, amplitude is saturated and, on defects higher than
6 mm, even reduced. As a result, large defects (8 mm and up) give an amplitude comparable to that of much
smaller defects (e.g. 3 mm). This effect is especially present when 60 and 70 probes are used on defects
originating from the gap (inside of the sleeve). This means that discrimination between small and very large
defects on the basis of corner effect alone will not be possible. This is shown in the graph of fig. 7, which is the
result of a series of simulations whereby defect height was increased in 0,2 mm increments and the amplitude

for each situation was stored. Simulated reference sensitivity was 80% screen height on a 5 mm high defect, in
order to keep the signals within the dynamic range. The top half of the figure shows intensities in the ultrasonic
beam.

amplitude height (% FSH)

notch depth (mm)

sound path (mm)

Fig. 7: Amplitude as a function of defect height (fusion defect, corner effect)

Reflection against the gap between sleeve and pipe wall, and subsequently against the weld cap and back,
generates a clear signal if no significant fusion defect originating from the gap is present. As soon as defect
height approaches 2 mm this geometric signal disappears. This signal could therefore be used as an additional
tool to discriminate between significant and non-significant fusion defects, even in manual UT. The same goes
for the transmission signal to the other side of the weld, which could be picked up on the pipe wall if a pitchcatch technique in mechanized UT would be used.
Fig. 8 shows a simulation on an underbead crack at the welds toe under an angle of 45, detected through the
weld using a 70 probe placed on the sleeve. The beam intensity is visualized. Fig. 9B shows, as an example,
the situation that the wave front has just passed the crystal after having been reflected on the defect (captured
image of moving wave front simulation). The A-scan (second echo, black) shows an amplitude that is approx. 6
dB lower than the reference echo (first signal, red), which is equivalent to 80% screen height on a 2 mm high
notch using corner effect (captured before at the optimum probe position for that particular reflector). In fig.
9A, the wave front is about to hit the defect. This type of simulations showed that this is a good way to detect
underbead cracks at the welds toe, without significant influence of the (varying) weld geometry as long as it is
within visual inspection acceptance limits. Sensitivity has to be slightly higher than 80% FSH on a 2 mm notch.

Fig. 8: Beam intensity on underbead defect

Fig. 9A: Wave front just about to hit defect

Fig. 9B: Reflected wave front just hit crystal

Round robin test results: On the thinner sleeves, fusion defects with a height of 4 mm and up were detected with
a PoD (Probability of Detection) of 92%. Length was, on average, overestimated by 23 mm, standard deviation of
length inaccuracy was +19%. On the thicker sleeves, defects with heights of 8 mm and up were detected with a
PoD of 80%, with a considerable length overestimation. From the results it became clear that the PoD decreases
with increasing sleeve thickness and defect height.
Apart from defect length overestimation, it became clear (from comparison to the fabrication and RT data) that UT
detected a significant amount of non-relevant welding-related imperfections. These appear at the extremity of the
gap between pipe wall and sleeve, mostly as a consequence of gas trapped between sleeve and pipe wall.
Discrimination between defects with significant heights and these small imperfections appeared to be far from
reliable. In practice, this could have resulted in many unnecessary repairs.
Defect sizing was fairly accurate for defects with heights up to 4 mm within 1.7 mm, compared to the fabrication
data of the intended defects. These practical results were in unison with the simulations.
Fig. 10 shows a typical example of the round robin test results on one of the welds.

Fig. 10: Example of round robin test results on 10 mm thick sleeve (weld B)

Legend:
Real defect position
Detected UT
Detected RT
(note: indicated defect height is shown as measured, except for RT results).
Discussion: When comparing the round robin results with the simulated results, the following can be concluded:
The sensitivity used in the current procedure for manual UT on perpendicular fusion defects is too high. For
fusion defects it is recommended to use a 2 mm notch rather than a 2 mm flat bottom hole (DGS) as a
reference. This is a sensitivity approx. 12 dB lower than the current sensitivity.
The round robin results confirm the simulated results in the sense that large defects can not be discriminated
from small, insignificant anomalies on the basis of corner effect amplitude alone.
The possibility of using geometrical signals as an additional tool for discrimination between large and small
defects was confirmed by the round robin results.
Another additional tool to discriminate large defects from small imperfections is, to have the complete area of
the defect contribute to the signal. This can be done by introducing tandem technique in a zonal concept on the
sleeve, using a phased array.
Underbead defects can reliably be detected through the weld, with the probe on the sleeve surface. Sensitivity
should be a 2 mm notch + 6 dB.
Consequences for future phase 2 work: At the time this paper was filed, phase 1 was completed. Phase 2 was in
its definition phase. Therefore, no phase 2 modeling results were available yet.
From phase 1 it became clear that there is far more information available in the ultrasonic signals than a single
manually scanned angle probe can pick up. The use of combinations of receiving angles, even tandem technique,
which is possible if phased array probes are used, is expected to increase POD as well as discrimination of large
from small defects. Based on the results of phase 1, phase 2 will include the following investigations directed to the
use of phased arrays:
Increase of the PoD of perpendicular fusion defects by using a combination of corner effect, tandem technique
(only practical when phased arrays are used) and pitch & catch.
Use of a combination of parameters to asses defect height:
- defect amplitude using a combination of corner effect and tandem technique
- use of information from the welds geometry
- use of pitch and catch method.
For detection of underbead cracks at the welds toe (of which the orientation is hard to predict) a combination
of pulse-echo from two sides will be modeled, using a variety of angles at shifting index points; in addition, it
will be investigated how the screening effect of an underbead defect on an ultrasonic beam (e.g. creeping
waves) can be used as an additional detection and sizing tool.
For detection of voluminous defects in the weld body, volumetric inspection using a phased array will be
modeled whereby the welds varying geometry is used as a reference rather than a disturbing factor.
Simulations with ToFD will include the use of multiple transmitters and receivers and optimization of their
beam characteristics and location, in order to arrive at an optimized combination of positive detection with
diffraction signals and screening of geometric signals (lateral wave, bottom reflection, mode converted
signals).

In Gasunie, a study is going on to establish specific acceptance criteria for the different types of defects that
can be expected in a repair sleeve fillet weld. These will have influence on the modeling parameters in the
phase 2 study.
Finally, a written procedure will be drafted that uses the optimized inspection parameters established in phase
2. With this procedure, the existing and additional fillet welds with artificial and natural defects will be
inspected, followed by radiographic and destructive examination.
Conclusions: Phase 1 computer modeling of ultrasonic testing of circumferential fillet welds of split-tee and repair
sleeves on pipelines has shown that, using manual UT, discrimination between relevant fusion defects and weldingrelated small, non-relevant imperfections is far from reliable. Large defects can give signals similar to those of
small imperfections.
This result was confirmed by a Round Robin test on four welds, using five operators.
Phase 1 modelling study showed that this performance can significantly be improved by making better use of
reflected and diffracted signals generated by defects and geometry. Although the current manual UT procedure
gives room for improvement, significant improvements can be expected from the use of mechanized techniques
using phased array and TOFD. This will be done in phase 2.
References:
1
Advances in Imaging of NDT Results; Niels Prtzgen, Dries Gisolf and Gerrit Blacquire;
paper no. 611, 16 WCNDT, Montreal, Canada, August 30 September 3, 2004